AI and Tech Innovation: A Roundup of Recent Developments

2023-11-10 10:34:48.028597

AI and Tech Innovation: A Roundup of Recent Developments

In a whirlwind of digital innovation, Meta Platforms is making a comeback in China, striking a deal with Tencent to sell a new, lower-priced VR headset, despite the global VR market taking a nosedive (source). Meanwhile, the recycling industry is getting a much-needed boost from AI, with robots sorting waste more efficiently than humans, despite the hefty upfront costs (source).

AI is also transforming work processes, expanding personal capabilities, and even giving some the feeling of having superpowers, despite the potential for misinformation (source). But not all is rosy in the AI world, with experts divided on the existential threat it poses, some warning of catastrophic harm, while others dismiss such fears as exaggerated (source).

The power-guzzling habits of AI are driving a search for alternative energy sources for data centers, with tech giants exploring wind, solar, geothermal, nuclear, and flared gas (source). In the semiconductor industry, innovation continues beyond Moore's Law, with AI potentially reducing chip design time from years to days (source).

Network carriers are exploring 'network slicing' for 5G, a technology that could conflict with net-neutrality regulations (source). And finally, in response to the growing issue of deepfake scam ads, US legislators are proposing bills to combat this menace, though the rapid evolution of deepfake technology could render these measures obsolete (source).

Meta Strikes Deal with Tencent to Sell VR Headsets in China

Meta Platforms has reached a preliminary agreement with Tencent Holdings to sell a new, lower-priced version of its virtual-reality (VR) headset in China, marking its return to the Chinese consumer market 14 years after Facebook was banned. Tencent, the world's largest video game company, will exclusively sell Meta's headsets in China from late 2024. The deal, which is still provisional and subject to change, could face regulatory hurdles as Beijing tightens regulations for digital content. The new headset will feature cheaper lenses than the Quest 3 model and a more advanced graphic processing unit than the Quest 2. Meta will take a larger share of device sales, while Tencent will earn more from content and service revenue. This move comes amid weakening global demand for VR headsets and skepticism about the technology's long-term potential. Meta was the global headset market leader in Q2 2023 with a 50% market share, but global headset shipments fell 44.6% from a year ago. The deal could potentially boost the sales and adoption of VR devices in China, despite the VR market experiencing a 56% decline in headset shipments in H1 2023.

AI Boosts Recycling Rates Amid Labor Shortages and Rising Costs

Recycling rates in the U.S. have been declining due to labor shortages and rising costs, but artificial intelligence (AI) is being deployed to reverse this trend. AI-driven robots and optical sorters are being used to sort recyclable waste more efficiently and at a faster rate than humans. The use of robotics in sorting centers has increased from less than 5% in 2019 to around 32% now. Waste Management, the largest waste manager in the U.S., is investing over $1 billion in recycling infrastructure, including AI and automation, aiming to increase its recovery of recyclable materials by 60% by 2030. Despite the upfront costs and maintenance requirements, AI is proving to be a cost-effective solution, with recyclers typically recouping their investment in robotic systems in two years. The data collected by AI-driven sorters is also helping states learn about the brands in their waste streams, aiding in the creation of extended producer responsibility programs.

AI Tools Transform Work Processes and Expand Personal Capabilities

In a Wall Street Journal article, technology researcher Alexandra Samuel shares her experiences and insights from using generative AI tools like ChatGPT, Scite, and Claude. She highlights how AI has transformed her work process, enabling rapid prototyping and validation of ideas, and expanding her capabilities beyond perceived limitations. Samuel emphasizes the importance of treating AI as a tutor rather than a tool to do the work, and the value of "prompt engineering" - a series of interactions with AI to get the best results. She also warns about the potential for AI to generate false information, stressing the need for fact-checking. Despite the challenges, Samuel describes the use of AI as empowering, likening it to gaining superpowers. This article underscores the transformative potential of AI in the workplace, but also highlights the need for careful and thoughtful engagement with these tools.

AI's Existential Threat: A Debate Among Experts

As artificial intelligence (AI) continues to advance, experts are divided on the potential risks it poses. Dan Hendrycks, director of the Center for AI Safety, warns that AI could lead to catastrophic harm, citing the potential for AI to enable nonexperts to bioengineer deadly pathogens and escalate cyberwarfare. He argues that current regulations are insufficient and that the rapid development of AI systems is outpacing safety measures. On the other hand, Arvind Narayanan, a professor of computer science at Princeton University, and Sayash Kapoor, a doctoral candidate at the university, argue that fears of AI are exaggerated. They contend that the idea of rogue AI is based on flawed science fiction concepts and that focusing on how future technology can cause harm ignores how it can help. They also argue that AI makes humans more powerful and that there is no reason to believe that AI acting alone will be more capable than humans aided by AI. Both sides agree that AI's potential risks need to be addressed, but they differ on the urgency and nature of those risks.

AI's Power-Guzzling Habits Drive Search for Alternative Energy Sources

The increasing energy demands of AI models are driving a search for alternative energy sources for data centers. AI could consume up to 3.5% of the world's electricity by 2030, according to Gartner. Tech giants like Amazon, Microsoft, and Google are exploring nontraditional energy sources, including wind, solar, geothermal, nuclear, and flared gas. Crusoe Energy Systems, a startup that converts flared gas into electricity for data centers, has over 100 small, modular data centers built on-site at oil wells. The company claims to have mitigated more methane than it has produced, reducing demand on power grids. However, critics argue that its strategy still relies on fossil fuels. Training one AI model like OpenAI’s GPT-3 can use up to 10 gigawatt-hours, equivalent to the electricity used by 1,000 U.S. homes in a year. Major cloud providers are also investing in renewable and alternative energy sources, with data centers accounting for 1% to 3% of global electricity use. The broader data center market has seen strong growth due to the AI boom, competition among cloud providers, and tighter supply. However, the International Energy Agency warns that existing efficiency improvements won't be enough to meet the growing size of AI models and demand for computing, predicting a net growth in AI-related energy use in the coming years.

Semiconductor Innovation Continues Beyond Moore's Law, Despite Challenges

Moore's Law, the principle that the number of transistors on a chip doubles roughly every two years, has been a guiding force in semiconductor advancement for decades. However, as transistor size has become increasingly small, the steady progress described by Moore's Law is reaching its limits. Despite this, companies like Nvidia and Google DeepMind are pushing forward with semiconductor innovation, particularly in the realm of artificial intelligence (AI). Nvidia's chief scientist, Bill Dally, believes that the path towards more innovation is clear for the next four years, with implications for sectors ranging from healthcare to e-commerce. Techniques such as sparsity and number representation are being explored to boost chip power and efficiency. Furthermore, AI is being employed to potentially reduce the time and cost of chip design, with some researchers believing it could reduce the design process from years to days. This could lead to massive productivity gains and transform the industry, according to Olivier Temam, director of data center and chip research at Google DeepMind.

Network Carriers Explore 'Network Slicing' for 5G, Despite Potential Net-Neutrality Conflicts

Network carriers AT&T and Verizon are exploring a technology called 'network slicing' to enhance the delivery of certain apps and services on their 5G networks. This technology would allow carriers to create a priority lane on the network for key services, potentially improving the performance of applications where speed is critical, such as telehealth, virtual reality, and gaming. However, this approach could conflict with net-neutrality regulations, which prohibit internet providers from discriminating among legal web traffic. Both AT&T and Verizon are currently testing network slicing and aim to deploy it in 2023. The carriers believe that network slicing could increase returns on their substantial investments in 5G upgrades. However, questions remain about how carriers will monetize this technology and whether it will conflict with evolving net-neutrality regulations. The Federal Communications Commission (FCC) has proposed reinstating an Obama-era policy that would restore net-neutrality mandates. AT&T and Verizon are working with regulators to ensure the appropriate implementation of network slicing.

US Legislators Propose Bills to Combat Deepfake Scam Ads

In response to the growing issue of AI-generated deepfake scam ads, US legislators have proposed bills to create a national standard prohibiting unauthorized deepfakes in a commercial context. The bills aim to protect both celebrities and ordinary citizens from scammers who use their likenesses in deceptive ads. However, the effectiveness of these measures against the rapid evolution of deepfake technology remains uncertain. Social media platforms have struggled to block the spread of deepfakes, with unauthorized ads featuring celebrities like Tom Hanks and MrBeast bypassing their filters. TikTok data suggests that 0.4% of the 106 million posts removed between April and June violated bans on synthetic or manipulated media. The proposed bills could impose fines of up to $5,000 per violation, with the House bill suggesting harsher penalties including civil judgments of up to $150,000 per violation and up to five years in prison for criminal violations. Despite these efforts, experts warn that the rapid advancement of deepfake technology could quickly render these bills obsolete.

None